Boosting the accuracy of differentially private histograms through consistency

Hay, Michael; Rastogi, Vibhor; Miklau, Gerome; Suciu, Dan

doi:10.14778/1920841.1920970

Cited by 461 publications

(585 citation statements)

References 20 publications

Supporting

Mentioning

583

Contrasting

Order By: Relevance

“…These counts are used to approximate the original counts in the database. By [6], this scheme is always consistent with 2 -differential privacy. Therefore, the complete modified algorithm of StructureFirst is still fulfilling -different privacy.…”

Section: B Structurefirst Methodsmentioning

confidence: 75%

“…To apply non-uniform scheme, we mainly borrow the idea from [6]. Generally speaking, after constructing all the bins, StructureFirst runs the boosting algorithm [6] is added on every count x j for l ≤ j ≤ r, as well as the sum s i = l≤j≤r x j . Assume x j is the result noisy count and s i is the noisy sum.…”

Section: B Structurefirst Methodsmentioning

confidence: 99%

“…There is some connection between such method and [6]. Our StructureFirst method with this non-uniform count assignment scheme can be considered as an adaptive version of the boosting algorithm in [6].…”

Section: B Structurefirst Methodsmentioning

confidence: 99%

“…Our StructureFirst method with this non-uniform count assignment scheme can be considered as an adaptive version of the boosting algorithm in [6]. In the original boosting algorithm, the user must specify the fan-out of the tree structure before running the algorithm.…”

Section: B Structurefirst Methodsmentioning

confidence: 99%

“…In particular, NoiseFirst can be used to improve the accuracy of an already published histogram using an existing method [3]. Moreover, we adapt DP-histograms to answer arbitrary range-count queries, which has drawn considerable research attention (e.g., [3], [6]). For such queries, NoiseFirst achieves better accuracy for short ranges, whereas StructureFirst is more suitable for longer ones.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Differentially Private Histogram Publication

Zhang

Xiao

et al. 2012

2012 IEEE 28th International Conference on Data Engineering

103

View full text Add to dashboard Cite

Abstract-Differential privacy (DP) is a promising scheme for releasing the results of statistical queries on sensitive data, with strong privacy guarantees against adversaries with arbitrary background knowledge. Existing studies on DP mostly focus on simple aggregations such as counts. This paper investigates the publication of DP-compliant histograms, which is an important analytical tool for showing the distribution of a random variable, e.g., hospital bill size for certain patients. Compared to simple aggregations whose results are purely numerical, a histogram query is inherently more complex, since it must also determine its structure, i.e., the ranges of the bins. As we demonstrate in the paper, a DP-compliant histogram with finer bins may actually lead to significantly lower accuracy than a coarser one, since the former requires stronger perturbations in order to satisfy DP. Moreover, the histogram structure itself may reveal sensitive information, which further complicates the problem.Motivated by this, we propose two novel algorithms, namely NoiseFirst and StructureFirst, for computing DP-compliant histograms. Their main difference lies in the relative order of the noise injection and the histogram structure computation steps. NoiseFirst has the additional benefit that it can improve the accuracy of an already published DP-complaint histogram computed using a naïve method. Going one step further, we extend both solutions to answer arbitrary range queries. Extensive experiments, using several real data sets, confirm that the proposed methods output highly accurate query answers, and consistently outperform existing competitors.

show abstract

Section: B Structurefirst Methodsmentioning

confidence: 75%

Section: B Structurefirst Methodsmentioning

confidence: 99%

Section: B Structurefirst Methodsmentioning

confidence: 99%

Section: B Structurefirst Methodsmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Differentially Private Histogram Publication

Zhang

Xiao

et al. 2012

2012 IEEE 28th International Conference on Data Engineering

103

View full text Add to dashboard Cite

show abstract

A multi‐dimensional index for privacy‐preserving queries in cloud computing

Ding

Jin

2019

Concurrency and Computation

View full text Add to dashboard Cite

Although cloud computing saves the cost of enterprises and individuals to manage data in various applications on public cloud servers, it also causes the problem of the leakage of critical personal information and sensitive data. Therefore, the issues about data privacy of the published personal information have become a challenging problem. Most of the existing methods focus on cryptography-based techniques by encrypting the sensitive data before publishing. However, these techniques are either too computation expensive or only some authorized users can get access to the encrypted cloud data. The consequence is that the search response time is not satisfactory and the sharing extent of the cloud data is becoming limited. Motivated by this, we propose a perturbation-based method to preserve the privacy of data in cloud. To accelerate the query processing without privacy breaches, a high-efficiency multi-dimensional index is built for answering range counting queries under differential privacy, which embeds privacy-preserving R-tree index in Content Addressable Network. Experimental results demonstrate that our methods not only protect data privacy in different degree, but also accelerate the query speed efficiently. Compared with the existing method Quad-opt, our method provides about 20% better data utility under the same differential privacy principles.

show abstract

Sensitivity reduction of degree histogram publication under node differential privacy via mean filtering

Lan

Huang

et al. 2019

Concurrency and Computation

View full text Add to dashboard Cite

Summary Publication of nodes' degree information in the form of histogram provides useful information about the graph as well as the risk of privacy disclosure. Under the robust protection of node‐differential privacy (node‐DP), publishing result's accuracy mainly depends on the global sensitivity of this publishing task. Thus, the reduction of sensitivity is of great importance. Existing methods for degree histogram publication under node‐DP are mostly based on limitation of maximum degree, whose sensitivity is still high, leading an unbearable noise scale. In this paper, we innovatively propose a method to tackle this issue. Firstly, we introduce mean filtering to process the histogram, almost halve the original sensitivity. Then, we use a series of techniques to further improve publishing accuracy, instituting a complete workflow for degree histogram publication under node‐DP. Experimental results show that our method effectively improves the accuracy.

show abstract

Boosting the accuracy of differentially private histograms through consistency

Cited by 461 publications

References 20 publications

Differentially Private Histogram Publication

Differentially Private Histogram Publication

A multi‐dimensional index for privacy‐preserving queries in cloud computing

Sensitivity reduction of degree histogram publication under node differential privacy via mean filtering

Contact Info

Product

Resources

About